Phase space sampling and inference from weighted events with autoregressive flows

We explore the use of autoregressive flows, a type generative model with tractable likelihood, as means efficient generation physical particle collider events. The usual maximum likelihood loss function is supplemented by an event weight, allowing for inference from samples variable, and even negative weights. To illustrate efficacy model, we perform experiments leading-order top pair production events at electron importance sampling weights, next-to-leading-order LHC that involve